Skip to Main Content
Scattered photons highly degrade the quality of X-ray images and their effect has become more important due to the increasing interest in cone-beam geometry for the acquisition of CT (CBCT) and micro-CT data. The random nature of scatter events and the great influence of the sample suggest that the most accurate methods for their estimation are Monte Carlo (MC) techniques, but their use is usually hampered by the large computation time required to obtain an acceptable estimation of the scattered radiation. We present an approach for scatter correction in CBCT by MC estimation, speeding up the computation by means of general purpose graphic processing units (GPGPU) and developing a framework for the automatic correction and reconstruction of projection data. The method consists of five stages: FDK reconstruction of the original data; histogram based automatic segmentation of the volume assigning a material and density to each voxel; fast MC estimation of the scatter signal; denoising of the independent scatter components and subtraction from original data; and FDK reconstruction of the corrected data. Every stage runs in a GPGPU using Nvidia CUD A. The MC stage is based on the MC-GPU code. To simulate polychromatic X-ray beams, the Spektr model is used to generate the source spectrum. Photon scattering is forced in order to reduce the number of events needed to obtain an acceptable scatter image weighting the photon histories to assure the correctness of the result. Further reduction in the variance is obtained by split the photon in several virtual photons which are forced point to the detector and are transported with no further interaction to the detector's surface. Furthermore, the divergence of the execution path of GPGPU kernels has been minimized. These techniques achieve a reduction of the variance of the scatter signal of two orders of magnitude and the final efficiency is improved by a factor of ~30.